(An Overview on) Linked Data Management and SPARQL Querying (ISSLOD2011)

An Overview on
Linked Data Management
and SPARQL Querying
Olaf Hartig
http://olafhartig.de/foaf.rdf#olaf
@olafhartig

Database and Information Systems Research Group
Humboldt-Universität zu Berlin

Outline

Part 1 Introduction to SPARQL

Part 2 Storage Solutions
for RDF Data

Part 3 Querying the
Web of Linked Data

Olaf Hartig - Linked Data Management and SPARQL Querying 2

Outline


Part 2 Storage do express
How Solutions
declarative queries
for RDF Data
over RDF data?
(A user perspective)

Part 3 Querying the
Web of Linked Data


SPARQL in General
● A family of W3C recommendations
● SPARQL Query
● Declarative query language for RDF data
● Our focus today
● SPARQL Update
● Declarative update language for RDF data
● SPARQL Protocol
● Communication between SPARQL processing
services (a.k.a. SPARQL endpoints) and clients
● SPARQL Query Results XML Format
● XML format for serializing query results

Main Idea of SPARQL Queries
● Pattern matching:
● Describe subgraphs of the queried RDF graph
● Subgraphs that match your description contribute an answer
● Mean: graph patterns (i.e. RDF graphs with variables)

?v rdf:type
umbel-sc:Volcano


Main Idea of SPARQL Queries
Queried RDF graph:

rdf:type
dbpedia:Mount_Baker umbel-sc:Volcano
p:lastEruption rdf:type
"1880" dbpedia:Mount_Etna

?v rdf:type
umbel-sc:Volcano
Result:
?v
dbpedia:Mount_Baker
dbpedia:Mount_Etna

Components of SPARQL Queries
PREFIX
PREFIX rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#>
rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#>
PREFIX
PREFIX umbel-sc: <http://umbel.org/umbel/sc/>
umbel-sc: <http://umbel.org/umbel/sc/>
SELECT ?v
SELECT ?v
FROM <http://example.org/myGeoData>
WHERE {
WHERE {
?v rdf:type umbel-sc:Volcano .
}
}
ORDER BY ?name
ORDER BY ?name

● Prologue:
● Prefix definitions for using compact URIs (CURIEs)
● Attention (difference to Turtle): No period (“.”) character as
separator

PREFIX
PREFIX
SELECT ?v
SELECT ?v
WHERE {
WHERE {
}
}
ORDER BY ?name
ORDER BY ?name

● Result form specification:
● SELECT, DESCRIBE, CONSTRUCT, or ASK
(more about that later)


PREFIX
PREFIX
SELECT ?v
SELECT ?v
WHERE {
WHERE {
}
}
ORDER BY ?name
ORDER BY ?name

● Dataset specification:
● Specify the RDF dataset to be queried


PREFIX
PREFIX
SELECT ?v
SELECT ?v
WHERE {
WHERE {
}
}
ORDER BY ?name
ORDER BY ?name

● Query Pattern:
● WHERE clause specifies the graph pattern to be matched


Graph Patterns

● Different types of graph patterns for the query pattern
(WHERE clause):
● Basic graph pattern (BGP)
● Group graph pattern
● Optional graph pattern
● Union graph pattern
● Graph graph pattern
● (Constraints)


Basic Graph Patterns
● Set of triple patterns (i.e. RDF triples with variables)
● Variable names prefixed with “?” or “$” (e.g. ?v, $v)
● Turtle syntax
● Including syntactical sugar (e.g. property and object lists)
PREFIX
PREFIX
PREFIX rdfs: <http://www.w3.org/2000/01/rdf-schema#>
rdfs: <http://www.w3.org/2000/01/rdf-schema#>
PREFIX
SELECT
SELECT ?name
?name
WHERE {
WHERE {
?v rdfs:label ?name .
?v rdfs:label ?name .
}
}

Basic Graph Patterns
● Set of triple patterns (i.e. RDF triples with variables)
● Variable names prefixed with “?” or “$” (e.g. ?v, $v)
● Turtle syntax
● Including syntactical sugar (e.g. property and object lists)
PREFIX
PREFIX
PREFIX rdfs: <http://www.w3.org/2000/01/rdf-schema#>
rdfs: <http://www.w3.org/2000/01/rdf-schema#>
PREFIX
SELECT
SELECT ?name
?name
WHERE {
WHERE {
?v rdf:type umbel-sc:Volcano ;
rdfs:label ?name .
rdfs:label ?name .
}
}

Basic Graph Patterns (Example)
dbpedia:Mount_Etna rdf:type umbel-sc:Volcano ;
dbpedia:Mount_Etna rdf:type umbel-sc:Volcano ; Data*
rdfs:label "Etna" .
rdfs:label "Etna" .
dbpedia:Mount_Baker rdf:type umbel-sc:Volcano.
dbpedia:Beerenberg rdf:type umbel-sc:Volcano,
umbel-sc:NaturalElevation ;
rdfs:label "Beerenberg"@en ;
rdfs:label "Бееренберг"@ru .
● Question: What are the names of all (known) volcanos?
SELECT ?name WHERE {
SELECT ?name WHERE { Query*
rdfs:label ?name . }

*Prefix definitions omitted

dbpedia:Mount_Etna rdf:type umbel-sc:Volcano ; Data*
rdfs:label "Etna" .
rdfs:label "Etna" .
●Question: What are the names of all (known) volcanos?
Result:
SELECT ?name WHERE { Query*
?name
"Etna"
"Бееренберг"@ru
*Prefix definitions omitted "Beerenberg"@en

dbpedia:Mount_Etna rdf:type umbel-sc:Volcano ; Data
rdfs:label "Etna" .
rdfs:label "Etna" .
● Question: List all types of the volcano called “Beerenberg”
SELECT ?type WHERE {
SELECT ?type WHERE { Query
?type
?v rdf:type ?type ;
?v rdf:type ?type ;
rdfs:label "Beerenberg" .
rdfs:label "Beerenberg" .
}
}
Empty!

rdfs:label "Etna" .
rdfs:label "Etna" .
● Question: List all types of the volcano called “Beerenberg”
SELECT ?type WHERE { Query
?v rdf:type ?type ;
?v rdf:type ?type ; ?type
rdfs:label "Beerenberg"@en .
rdfs:label "Beerenberg"@en . umbel-sc:Volcano
}
} umbel-sc:NaturalElevation

dbpedia:Mount_Baker rdf:type umbel-sc:Volcano ;
dbpedia:Mount_Baker rdf:type umbel-sc:Volcano ; Data
p:location dbpedia:United_States .
dbpedia:United_States rdfs:label "United States" .
dbpedia:United_States rdfs:label "United States" .

● Question: Where are all (known) volcanos located?
(List the names of these locations)
● Blank nodes in SPARQL queries
● As subject or object of a triple pattern
● “Non-selectable” variables

SELECT ?name WHERE { Query
?name
_:x rdf:type umbel-sc:Volcano ;
_:x rdf:type umbel-sc:Volcano ;
"United States"
p:location [ rdfs:label ?name ] . }
p:location [ rdfs:label ?name ] . }

dbpedia:Mount_Baker rdf:type umbel-sc:Volcano ; Data
p:location [ rdfs:label "United States"@en ,
p:location [ rdfs:label "United States"@en ,
"États-Unis"@fr ] .
"États-Unis"@fr ] .
p:location [ rdfs:label "Italy" ] .
p:location [ rdfs:label "Italy" ] .

● Blank nodes in the queried data
➔ Blank node identifiers may occur in the results
SELECT ?l ?name WHERE {
SELECT ?l ?name WHERE { Query
?v rdf:type umbel-sc:Volcano ; ?l ?name
p:location
p:location ?l .
?l . _:x "United States"@en
?l rdfs:label
?l rdfs:label ?name .
?name . _:x "États-Unis"@fr
} _:y "Italy"
}

Optional Graph Patterns
rdfs:label "Etna" .
rdfs:label "Etna" .
dbpedia:Mount_Baker rdf:type umbel-sc:Volcano .
dbpedia:Beerenberg rdf:type umbel-sc:Volcano ;

● Question: What are all (known) volcanos and their names?
SELECT ?v ?name WHERE {
SELECT ?v ?name WHERE { Query
?v ?name
● Problem: Mount
dbpedia:Mount_Etna "Etna"
Baker missing
dbpedia:Beerenberg "Beerenberg"@en
(it has no name)

Optional Graph Patterns
● Keyword OPTIONAL allows for optional patterns
SELECT ?v ?name WHERE {
SELECT ?v ?name WHERE { Query
OPTIONAL { ?v rdfs:label ?name }
}
}

?v ?name
dbpedia:Mount_Baker
● Optional patterns may result in unbound variables


Union Graph Patterns
rdfs:label "Etna" ;
rdfs:label "Etna" ;
p:location dbpedia:Italy .
rdfs:label
"Beerenberg"@en ;
p:location
p:location dbpedia:Norway .
dbpedia:Norway .

● Question: What volcanos are located in Italy or in Norway?
SELECT ?v WHERE {
SELECT ?v WHERE { Query
p:location
p:location ?
? . }
. }

● Union graph patterns allow for alternatives
SELECT ?v WHERE { ?v rdf:type umbel-sc:Volcano .
{ ?v p:location dbpedia:Italy }
UNION
UNION
{ ?v p:location dbpedia:Norway }
}
} Query


● Union graph patterns allow for alternatives
UNION
UNION
}
} Query
SELECT
SELECT ?v WHERE
?v WHERE {
{ Query
{ ?v
{ ?v rdf:type
rdf:type umbel-sc:Volcano ;
umbel-sc:Volcano ;
p:location dbpedia:Italy }
p:location dbpedia:Italy }
Semantically UNION
UNION
equivalent { ?v rdf:type umbel-sc:Volcano
{ ?v rdf:type umbel-sc:Volcano ;
;
p:location dbpedia:Norway
p:location dbpedia:Norway }
}
}
}

Group Graph Patterns
UNION
UNION
}
} Query

Semantically equivalent

SELECT ?v WHERE { { ?v rdf:type umbel-sc:Volcano }
SELECT ?v WHERE { { ?v rdf:type umbel-sc:Volcano }
{ { ?v p:location dbpedia:Italy }
{ { ?v p:location dbpedia:Italy }
UNION
UNION
{ ?v p:location dbpedia:Norway } }
{ ?v p:location dbpedia:Norway } }
}
} Query

Constraints on Solutions
● Syntax: Keyword FILTER followed by filter expression
PREFIX
PREFIX
PREFIX p: <http://dbpedia.org/property/>
PREFIX p: <http://dbpedia.org/property/>
SELECT ?v
SELECT ?v
WHERE {
WHERE {
p:lastEruption ?le .
p:lastEruption ?le .
FILTER ( ?le > 1900 )
FILTER ( ?le > 1900 )
}
}

● Filter expressions contain operators and functions
● Operators and functions operate on RDF terms

Constraints (Truth Table)
● Filter expressions evaluate to true, false, or error
● Truth table:
A B A || B A && B
T T T T
T F T F
F T T F
F F F F
T E T E
E T T E
F E E F
E F E F
E E E E


Unary Operators in Constraints

Operator Type(A) Result type
!A xsd:boolean xsd:boolean
+A numeric numeric
-A numeric numeric
BOUND(A) variable xsd:boolean
isURI(A) RDF term xsd:boolean
isBLANK(A) RDF term xsd:boolean
isLITERAL(A) RDF term xsd:boolean
STR(A) literal / URI simple literal
LANG(A) literal simple literal
DATATYPE(A) literal simple literal


Constraints (Example)
rdfs:label "Etna" .
rdfs:label "Etna" .
●Question: List all types of the volcano called “Beerenberg”
SELECT ?type WHERE { Query ?type
?v rdf:type ?type ;
?v rdf:type ?type ; umbel-sc:Volcano
rdfs:label ?name .
rdfs:label ?name .
umbel-sc:NaturalElevation
FILTER ( STR(?name) = "Beerenberg" )
FILTER ( STR(?name) = "Beerenberg" )
}
}


Constraints (Further Operators)
● Binary operators:
● Logical connectives && and || for xsd:boolean
● Comparison operators =, !=, <, >, <=, and >= for numeric
datatypes, xsd:dateTime, xsd:string, and xsd:boolean
● Comparison operators = and != for other datatypes
● Arithmetic operators +, -, *, and / for numeric datatypes

● Furthermore:
● REGEX(String,Pattern) or REGEX(String,Pattern,Flags)
● sameTERM(A,B)
● langMATCHES(A,B)


rdfs:label "Etna" .
rdfs:label "Etna" .
● Question: What volcanos have an “e” in their name?
SELECT ?v WHERE {
?v
rdfs:label ?name . dbpedia:Beerenberg
rdfs:label ?name .
dbpedia:Beerenberg
FILTER( REGEX(STR(?name),"e") )
FILTER( REGEX(STR(?name),"e") )
}
}


rdfs:label "Etna" .
rdfs:label "Etna" .
● Question: What volcanos have an “e” in their name?
SELECT ?v WHERE {
?v
rdfs:label ?name . dbpedia:Mount_Etna
rdfs:label ?name .
dbpedia:Beerenberg
FILTER( REGEX(STR(?name),"e","i") )
FILTER( REGEX(STR(?name),"e","i") ) dbpedia:Beerenberg
}
}


Graph Graph Patterns
dbpedia:Mount_Etna rdfs:seeAlso <http://example.org/d1>.
dbpedia:Mount_Baker rdfs:seeAlso <http://example.org/d2>.
dbpedia:Mount_Etna
dbpedia:Mount_Etna http://example.org/d1
Default
rdfs:label "Etna" .
rdfs:label "Etna" .
Graph
dbpedia:Mount_Baker http://example.org/d2
dbpedia:Mount_Baker
rdf:type umbel-sc:Volcano .
dbpedia:Beerenberg http://example.org/d3
dbpedia:Beerenberg
● SPARQL queries rdf:type umbel-sc:Volcano ;
are executed over rdfs:label "Beerenberg"@en .
an RDF dataset:
● One default graph and
● Zero or more named graphs (identified by an URI)

PREFIX
PREFIX
SELECT ?v
SELECT ?v
WHERE {
WHERE {
}
}
ORDER BY ?name
ORDER BY ?name

● Dataset specification:
● Specify the RDF dataset to be queried
(more about that later) here ...
● Specification using FROM and FROM NAMED

dbpedia:Mount_Etna
Default
rdfs:label "Etna" .
rdfs:label "Etna" .
Graph
dbpedia:Mount_Baker
dbpedia:Beerenberg

● Evaluation of patterns w.r.t. active graph
● GRAPH clause for making a named graph the active graph

dbpedia:Mount_Etna
Default
rdfs:label "Etna" .
rdfs:label "Etna" .
Graph
dbpedia:Mount_Baker
SELECT ?v WHERE { dbpedia:Beerenberg
SELECT ?v WHERE {
GRAPH <http://example.org/d1> umbel-sc:Volcano ;
rdf:type
GRAPH <http://example.org/d1> {{
}
} ?v
}
} dbpedia:Mount_Etna

dbpedia:Mount_Etna
Default
rdfs:label "Etna" .
rdfs:label "Etna" .
Graph
dbpedia:Mount_Baker
SELECT ?v WHERE { dbpedia:Beerenberg
SELECT ?v WHERE {
GRAPH ?g {
GRAPH ?g { ?v
?v rdf:type umbel-sc:Volcano . dbpedia:Mount_Etna
}
} dbpedia:Mount_Baker
}
} dbpedia:Beerenberg

dbpedia:Mount_Etna
Default
rdfs:label "Etna" .
rdfs:label "Etna" .
Graph
dbpedia:Mount_Baker
dbpedia:Beerenberg
SELECT ?v ?g WHERE {
SELECT ?v ?g WHERE {
GRAPH ?g {
GRAPH ?g { ?v ?g
?v rdf:type umbel-sc:Volcano . <http://example.org/d1>
dbpedia:Mount_Etna
}
} dbpedia:Mount_Baker <http://example.org/d2>
}
} dbpedia:Beerenberg <http://example.org/d3>

dbpedia:Mount_Etna
Default
rdfs:label "Etna" .
rdfs:label "Etna" .
Graph
dbpedia:Mount_Baker
SELECT ?v WHERE {
SELECT ?v WHERE {
dbpedia:Beerenberg
_:x rdfs:seeAlso ?g
_:x rdfs:seeAlso ?g
GRAPH ?g {
GRAPH ?g {
?v rdf:type umbel-sc:Volcano . ?v
}
}
dbpedia:Mount_Etna
}
} dbpedia:Mount_Baker

Negation
rdfs:label "Etna" .
rdfs:label "Etna" .
rdfs:label
rdfs:label "Beerenberg"@en
"Beerenberg"@en ;
;
rdfs:label
rdfs:label "Бееренберг"@ru
"Бееренберг"@ru .
.
● Question: What volcanos do not have a name in our data?
SELECT ?v WHERE {
OPTIONAL { ?v rdfs:label ?name } ?v
FILTER( ! BOUND(?name) )
FILTER( ! BOUND(?name) ) dbpedia:Mount_Baker
}
}
Negation as Failure

Negation
rdfs:label "Etna" .
rdfs:label "Etna" .
rdfs:label
"Beerenberg"@en ;
;
rdfs:label
.
● Question: What volcanos are not called “Beerenberg”?
SELECT ?v WHERE {
?v rdf:type umbel-sc:Volcano . !=
rdfs:label ?name .
rdfs:label ?name .
FILTER (STR(?name) != "Beerenberg") ?v
FILTER (STR(?name) != "Beerenberg")
}
}
dbpedia:Mount_Etna
dbpedia:Mount_Baker
Olaf Hartig - Linked Data Management and SPARQL Querying dbpedia:Beerenberg 41

Negation
rdfs:label "Etna" .
rdfs:label "Etna" .
rdfs:label
"Beerenberg"@en ;
;
rdfs:label
.
●Question: What volcanos are not called “Beerenberg”?
SELECT ?v WHERE {
SELECT ?v WHERE {
?v
Query
?v rdf:type umbel-sc:Volcano . dbpedia:Mount_Etna
OPTIONAL { ?v rdfs:label ?name .
OPTIONAL { ?v rdfs:label ?name .
dbpedia:Mount_Baker
FILTER (STR(?name) = "Beerenberg") }
FILTER (STR(?name) = "Beerenberg") }
FILTER ( ! BOUND(?name) )
FILTER ( ! BOUND(?name) )
} Negation as Failure
}

dbpedia:Mount_Etna
Default
rdfs:label "Etna" .
rdfs:label "Etna" .
Graph
dbpedia:Mount_Baker
dbpedia:Beerenberg
● Question: Which named graphs contain the name of a
volcano that is not referenced in the default graph?

dbpedia:Mount_Etna
Default
rdfs:label "Etna" .
rdfs:label "Etna" .
Graph
SELECT ?g WHERE
SELECT ?g WHERE {
{dbpedia:Mount_Baker
GRAPH ?g {
GRAPH ?g { rdf:type umbel-sc:Volcano .
dbpedia:Beerenberg
rdfs:label ?name .
rdfs:label ?name .
}}
OPTIONAL { ?v rdfs:seeAlso ?r } "Beerenberg"@en .
rdfs:label
OPTIONAL { ?v rdfs:seeAlso ?r }
● Question: ! BOUND(?r) ) graphs contain the name of a
FILTER ( Which named
FILTER ( ! BOUND(?r) )
}
} volcano that is not referenced in the default graph?

Summary – Graph Patterns

● Different types of graph patterns for the query pattern
(WHERE clause):
● Basic graph pattern (BGP)
● Group graph pattern
● Optional graph pattern – keyword OPTIONAL
● Union graph pattern – keyword UNION
● Graph graph pattern – keyword GRAPH
● Constraints – keyword FILTER


PREFIX
PREFIX
SELECT ?v
SELECT ?v
WHERE {
WHERE {
}
}
ORDER BY ?name
ORDER BY ?name

● Result form specification:
● SELECT, DESCRIBE, CONSTRUCT, or ASK
^ here ...


Result Forms
● SELECT
● Result: sequence of solutions (i.e. sets of variable bindings)
● Selected variables separated by space (not by comma!)
● Asterisk character (“*”) selects all variables in the pattern
● ASK
● Check whether there is at least one result
● Result: true or false
● Example: Do we have data about volcanos?
ASK WHERE {
ASK WHERE { Query
}
}

Result Forms
● DESCRIBE
● Result: an RDF graph with data about resources
● Non-deterministic (i.e. query processor defines the actual
structure of the resulting RDF graph)
● Example: just name the resource
DESCRIBE <http://dbpedia.org/resource/Beerenberg>
DESCRIBE <http://dbpedia.org/resource/Beerenberg> Query
● Example: Specify the resource(s) with a query pattern
DESCRIBE ?v WHERE {
DESCRIBE ?v WHERE { Query
rdfs:label ?name .
}
● Multiple variables possible or asterisk (“*”) for all

Result Forms
● CONSTRUCT
● Result: an RDF graph constructed from a template
● Template: graph pattern with variables from the query pattern
CONSTRUCT { ?v rdfs:label ?name ;
CONSTRUCT { ?v rdfs:label ?name ; Query
rdf:type myTypes:VolcanosOutsideTheUS
rdf:type myTypes:VolcanosOutsideTheUS
}
}
WHERE {
WHERE {
rdfs:label ?name .
rdfs:label ?name .
OPTIONAL { ?v p:location
OPTIONAL { ?v p:location ?l
?l
FILTER ( ?l =
FILTER ( ?l = dbpedia:United_States ) }
dbpedia:United_States ) }
FILTER ( ! BOUND(?l) )
FILTER ( ! BOUND(?l) )
}
}

Result Forms
rdfs:label "Etna" ;
rdfs:label "Etna" ;
rdfs:label "Mount Baker" ;
rdfs:label "Mount Baker" ;

dbpedia:Mount_Etna rdfs:label "Etna" ;
dbpedia:Mount_Etna rdfs:label "Etna" ; Result
rdf:type myTypes:VolcanosOutsideTheUS.
rdf:type myTypes:VolcanosOutsideTheUS.
dbpedia:Beerenberg rdf:type myTypes:VolcanosOutsideTheUS;
dbpedia:Beerenberg rdf:type myTypes:VolcanosOutsideTheUS;

PREFIX
PREFIX
SELECT ?v
SELECT ?v
WHERE {
WHERE {
}
}
ORDER BY ?name
ORDER BY ?name

● Solution modifiers:
● Only for SELECT queries
● Modify the result set as a whole (not single solutions)
● Keywords: DISTINCT, ORDER BY, LIMIT, and OFFSET

Solution Modifiers
● DISTINCT removes duplicates from the result set
rdfs:label "Etna" .
rdfs:label "Etna" .
?type
SELECT ?type
SELECT ?type Query
umbel-sc:Volcano
WHERE { _:x rdf:type ?type }
WHERE { _:x rdf:type ?type } umbel-sc:Volcano
umbel-sc:NaturalElevation
umbel-sc:Volcano

Solution Modifiers
● DISTINCT removes duplicates from the result set
rdfs:label "Etna" .
rdfs:label "Etna" .
?type
SELECT DISTINCT ?type
SELECT DISTINCT ?type Query
umbel-sc:Volcano
WHERE { _:x rdf:type ?type }
WHERE { _:x rdf:type ?type } umbel-sc:NaturalElevation


Solution Modifiers
● ORDER BY orders the results
SELECT ?v WHERE { ?v rdf:type umbel-sc:Volcano ;
SELECT ?v WHERE { ?v rdf:type umbel-sc:Volcano ; Query
rdfs:label ?name }
rdfs:label ?name }
ORDER BY ?name
ORDER BY ?name
● How do we order different kinds of elements?
unbound variable < blank node < URI < literal
● ASC for ascending (default) and DESC for descending
● Hierarchical order criteria:
SELECT ?name WHERE { ?v rdf:type umbel-sc:Volcano ; Query
SELECT ?name WHERE { ?v rdf:type umbel-sc:Volcano ;
p:lastEruption ?le ;
p:lastEruption ?le ;
rdfs:label ?name }
rdfs:label ?name }
ORDER BY DESC(?le), ?name
ORDER BY DESC(?le), ?name

Solution Modifiers
● LIMIT – limits the number of results
rdfs:label ?name }
rdfs:label ?name }
ORDER
ORDER BY ?name
BY ?name
LIMIT
LIMIT 5
5
● OFFSET – position/index of the first reported results
rdfs:label ?name }
rdfs:label ?name }
ORDER
ORDER BY ?name
BY ?name
LIMIT
LIMIT 5 OFFSET 10
5 OFFSET 10
● Order of result should be predictable
(i.e. combine with ORDER BY)

SPARQL 1.1
● New features of SPARQL 1.1 Query:
● Aggregate functions (e.g. COUNT, SUM, AVG)
● Subqueries
● Negation (EXISTS, NOT EXISTS, MINUS)
● Assignments (e.g. BIND, SELECT expressions)
● Property paths
● Basic query federation (SERVICE, BINDINGS)
● SPARQL 1.1 Update:
● Graph update (INSERT DATA, DELETE DATA, INSERT,
DELETE, DELETE WHERE, LOAD, CLEAR)
● Graph management (CREATE, DROP, COPY, MOVE, ADD)


Further Reading
● W3C SPARQL Working Group
http://www.w3.org/2001/sw/DataAccess/
● J. Pérez, M. Arenas, and C. Gutierrez: Semantics and
Complexity of SPARQL. ACM Transactions on Database
Systems 34(3), September 2009
● SPARQL interface
for DBpedia:
http://dbpedia.org/snorql/


Outline


for RDF Data

Part 3 Querying the
Web of Linked Data


Outline


for RDF Data
How can we manage RDF data in
Part 3 Querying the an efficient and scalable way?
Web of (relational) DBMS, or
Use existing Linked Data
●

● Create a new, native DBMS for RDF


Outline


for RDF Data
(RDF in a RDBMS)
Part 3 Querying the
Web of Linked Data


Overview on RDF in RDBMS
● Idea: Use a specific relational schema for RDF data and
benefit from 40 years of research in the DB community
● 3 steps for SPARQL query processing:
1. Convert SPARQL query to SQL query (w.r.t. the schema)
2. Use RDBMS to answer the SQL query S P O
3. Generate SPARQL query result from
the SQL query result
S P1 ... Pi ... Pn
● Relational schemas for RDF:
● Triple table
● Property tables P1 ... Pi ... Pn
● Binary tables S O S O S O
● Hybrid

Triple Table (Basic Idea)
● Store all S P O
RDF triples
in a single
table


Triple Table (Basic Idea)
● Store all S P O
RDF triples dbpedia:Mount_Etna rdf:type umbel-sc:Volcano
in a single dbpedia:Mount_Etna rdfs:label "Etna"
table dbpedia:Mount_Baker rdf:type umbel-sc:Volcano
dbpedia:Beerenberg rdf:type umbel-sc:Volcano
dbpedia:Beerenberg rdfs:label "Beerenberg"@en

● Create indexes on combinations of S, P, and O

dbpedia:Mount_Etna rdf:type umbel-sc:Volcano ; RDF Data
rdfs:label "Etna" .
rdfs:label "Etna" .

Triple Table (Querying)
● Store all S P O
RDF triples
dbpedia:Mount_Etna rdf:type umbel-sc:Volcano
in a single
dbpedia:Mount_Etna rdfs:label "Etna"
table dbpedia:Mount_Baker rdf:type umbel-sc:Volcano
SELECT ?name WHERE { SPARQL Query
dbpedia:Beerenberg
SELECT ?name WHERE { rdfs:label "Beerenberg"@en
● Create indexes on combinations of S, P, and O
rdfs:label ?name .
rdfs:label ?name .
}
} SELECT umbel-sc:Volcano ; SQL Query
SELECT t2.o AS name
dbpedia:Mount_Etna rdf:type t2.o AS name
FROM triples AS t1, triples AS t2
FROM triples AS t1, triples AS t2
rdfs:label "Etna" .
rdfs:label "Etna" .
WHERE t1.p = 'rdf:type' AND
dbpedia:Mount_Baker rdf:type umbel-sc:VolcanoAND
WHERE t1.p = 'rdf:type' .
SPARQL t1.o = "umbel-sc:Volcano" AND
dbpedia:Beerenberg rdf:typet1.o = "umbel-sc:Volcano" AND
to SQL umbel-sc:Volcano ;
t2.p = 'rdfs:label' AND
rdfs:label "Beerenberg"@en .AND
t2.p = 'rdfs:label'
t1.s = t2.s
t1.s = t2.s

ID-based Triple Table
S P O
dbpedia:Mount_Etna rdf:type umbel-sc:Volcano
dbpedia:Mount_Etna rdfs:label "Etna"
dbpedia:Mount_Baker rdf:type umbel-sc:Volcano
RDF Term dbpedia:Beerenberg
ID rdfs:label "Beerenberg"@en

● Use numerical identifier for
each RDF term in the dataset
● Advantages:
● Saves space
● Enhances efficiency


ID-based Triple Table
S P O
1 4 6
1 5 7
2 4 6
3 4 6
RDF Term ID 3 5 8
dbpedia:Mount_Etna 1
dbpedia:Mount_Baker 2 ● Use numerical identifier for
dbpedia:Beerenberg 3 each RDF term in the dataset
rdf:type 4 ● Advantages:
rdfs:label 5
● Saves space
umbel-sc:Volcano 6
"Etna" 7
● Enhances efficiency
"Beerenberg"@en 8

Quad Table
G S P O
100 1 4 6
100 1 5 7
101 2 4 6
102 3 4 6
RDF Term
100 ID 3 5 8
dbpedia:Mount_Etna 1
dbpedia:Mount_Baker 2 ● Storing multiple RDF graphs
dbpedia:Beerenberg 3 ● Use Cases:
rdf:type 4
rdfs:label 5
● Provenance
umbel-sc:Volcano 6 ● Versioning (multiple snapshots)
"Etna" 7 ● Contexts
"Beerenberg"@en 8

Pros and Cons of Triple Tables

● Pros:
● Easy to implement
● If predicates are selective, not expensive
● Cons:
● Many self joins are needed
● If predicates are not selective, very expensive


Property Tables (Basic Idea)
● Combining all (or some) properties
of similar subjects in n-ary tables
S rdf:type rdfs:label
dbpedia:Mount_Etna umbel-sc:Volcano "Etna"
dbpedia:Mount_Baker umbel-sc:Volcano NULL
dbpedia:Beerenberg umbel-sc:Volcano "Beerenberg"@en
● Also uses ID based encoding

rdfs:label "Etna" .
rdfs:label "Etna" .

Pros and Cons of Property Tables

● Pros:
● Fewer joins
● If the data is structured, we have a relational DB
● Cons:
● Potentially a lot of NULLs
● Clustering is not trivial
● Multi-value properties are complicated


Binary Tables (Basic Idea)
● Also called vertical partitioning
● n two column tables (n is the number of unique
properties in the data)
● Also combined with rdf:type
ID based encoding S O
of RDF terms dbpedia:Mount_Etna umbel-sc:Volcano
dbpedia:Mount_Baker umbel-sc:Volcano
dbpedia:Beerenberg umbel-sc:Volcano

rdfs:label
S O

Row Store vs. Column Store
● Different physical storage models for relational DBs
● Row based storage:
● Tuples (i.e. DB records)
are stored consecutively
● Entire row needs to be
read even if few attributes
are projected
● Column based storage:
● Read columns relevant to the 1 2 1 2 1 2

query → projection is free 3 3 3

● Inserts are expensive
● Column based storage model fit binary tables very well

Performance Comparison

Copied from:
D.J. Abadi, A. Marcus, S.R.
Madden, and K. Hollenbach.
Scalable Semantic Web
Data Management Using
Vertical Partitioning. In Proc.
of VLDB 2007

● However: property tables may outperform triple table and
binary tables for RDF data with regular structure*
*J.J. Levandoski and M.F. Mokbel. RDF data-centric storage. ICWS 2009

Pros and Cons of Binary Tables
● Pros:
● Supports multi-value properties
● No NULLs
● Read only needed attributes (i.e. less I/O)
● No clustering
● Excellent performance (if number of properties
is small, queries with bounded properties)
● Cons:
● Expensive inserts
● Bad performance (large number of properties,
queries with unbounded properties)


Summary (RDF in RDBMS)
● Relational schemas for RDF: S P O
● Triple table S P1 ... Pi ... Pn
● Property tables
● Binary tables
● Hybrid P1 ... Pi ... Pn
S O S O S O
● None of these approaches
can compete with a purely
relational model:
“... there is still room for optimization in the proposed generic
relational RDF storage schemes and thus new techniques for
storing and querying RDF data are still required ...”*
* S. Sakr and G. Al-Naymat. Relational Processing of
RDF Queries: A Survey. SIGMOD Record 38(4), 2009

Outline


for RDF Data

Part 3 Querying the
Web of Linked Data


Outline

Interesting …
but, that was only about (local) RDF data.
Part 2 Storage Solutions Web?
What's about Linked Data on the
for RDF Data

Part 3 Querying the
Web of Linked Data


Can we query the
Web of Linked Data
as if it were a single
(gigantic) database?

SELECT DISTINCT ?i ?label
WHERE {

?prof rdf:type <http://res ... data/dbprofs#DBProfessor> ;
foaf:topic_interest ?i .

}
OPTIONAL {

}
?i rdfs:label ?label
FILTER( LANG(?label)="en" || LANG(?label)="")

ORDER BY ?label
?

Olaf Hartig - The Impact of Data Caching on Query Execution for Linked Data 78

Warehousing Based Approach
Traditional approach 1:
Data centralization (e.g. data warehouse)
● Querying a collection of
copies from all relevant
datasets

+ Most efficient
– Misses unknown or new sources
– Collection possibly out of date
Olaf Hartig - A Novel Approach to Query the Web of Linked Data

Query Federation Approach
Traditional approach 2:
Federated query processing ?
● Mediator distributes subqueries
to relevant sources and
integrates the results

?
+ Current data ? ?
– Sources must
provide a
query service
– Misses unknown
or new sources

Main Drawback

You have to know the relevant
data sources in advance.

You restrict yourself to
the selected sources.
You do not tap the
full potential of
the Web !


A Novel Approach

*
Link Traversal Based Query Execution

* First Publication: O. Hartig, C. Bizer, and J.-C. Freytag. Executing SPARQL
Queries over the Web of Linked Data. In Proc. of ISWC 2009

Main Idea
● Intertwine query evaluation with traversal of data links
● We alternate between:
● Evaluate parts of the query (triple patterns)
on a continuously augmented set of data
● Look up URIs in intermediate
solutions and add retrieved data
to the query-local dataset

Queried data

Main Idea

Query
http://.../movie2449
film
ing
n
r_i

Lo
ca
to

t io
ac

n
lives_in
?actor ?loc
Queried data

Main Idea
We alternate between:

htt
●

p:/
/.
Evaluate parts of the query (triple patterns)

../m ?
●


ov
ie2
44

9

Query
film
ing
n
r_i

Lo
ca
to

t io
ac

n
lives_in
?actor ?loc
Queried data

Main Idea
Query
http://.../movie2449 in
t or_
film http://mdb.../Paul ac
ing
n
r_i

Lo
ca
to

t io
ac

n
lives_in
?actor ?loc
Queried data

Main Idea
?actor
http://mdb.../Paul
Query
http://.../movie2449 in
t or_
film http://mdb.../Paul ac
ing
n
r_i

Lo
ca
to

t io
ac

n
lives_in
?actor ?loc
Queried data

Main Idea
?actor
http://mdb.../Paul

? aul
P

.../
db
/m
Look up URIs in intermediate

p:/
●

htt

Query
film
ing
n
r_i

Lo
ca
to

t io
ac

n
lives_in
?actor ?loc
Queried data

Main Idea
?actor
http://mdb.../Paul

Query
film
ing
n
r_i

Lo
ca
to

t io
ac

n
lives_in
?actor ?loc
Queried data

(An Overview on) Linked Data Management and SPARQL Querying (ISSLOD2011)

(An Overview on) Linked Data Management and SPARQL Querying (ISSLOD2011)

Recommended

Recommended

More Related Content

What's hot

What's hot (20)

Viewers also liked

Viewers also liked (16)

Similar to (An Overview on) Linked Data Management and SPARQL Querying (ISSLOD2011)

Similar to (An Overview on) Linked Data Management and SPARQL Querying (ISSLOD2011) (20)

More from Olaf Hartig

More from Olaf Hartig (14)

Recently uploaded

Recently uploaded (20)

(An Overview on) Linked Data Management and SPARQL Querying (ISSLOD2011)